Soil Dynamics and Earthquake Engineering 101 (2017) 105–115

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

journal homepage: www.elsevier.com/locate/soildyn

Numerical modeling of the interaction of pressurized large diameter gas MARK

buried pipelines with normal fault ruptures
⁎
AG Özcebea,b, , R. Paoluccia, S. Mariania
a
Department of Civil and Environmental Engineering, Politecnico di Milano, Milano, Italy
b
European Centre for Training and Research in Earthquake Engineering (EUCENTRE), Pavia, Italy

A R T I C L E I N F O A B S T R A C T

Keywords: There is a growing attention towards the seismic response of large diameter pipelines, owing to the potential
Fault-soil-pipeline interaction adverse impact on economy and civilized life of a structural collapse under earthquake effects, such as strong
3D finite element modeling ground shaking, and other earthquake related effects such as fault rupture, landsliding and liquefaction. The
Performance based seismic design intersection of a fault rupture with a pipeline is of special concern, because the safety verification is affected by
significant uncertainties in the loading condition, related to the unknown exact location where the fault offset
may occur, the unknown amount of the offset itself, as well as the intersection angle of the fault rupture with the
pipe axis. Besides, the inherent analytical/numerical complexity of the problem may require 3D finite element
models with non-linear constitutive laws and large deformations.
In this paper a summary is presented of a comprehensive set of 3D numerical simulations of the interaction of
a large diameter gas pipeline with a normal fault rupture, with the main objectives of: 1) throwing light on the
pipeline performance under increasing levels of fault offset, including cross-sectional buckling and ovalization;
2) providing a parametric set of results, including the variability of the fault-pipe intersection angles, of the
mechanical properties of the pipe-soil interface, as well as of the operating conditions, in terms of internal gas
pressure and temperature variations.

1. Introduction
Seismic response of buried pipelines is mainly governed by the ki-
nematic interaction between soil and structure rather than by inertial
actions [1]. On one side, such interaction may induce transient effects,
related to straining due to seismic wave propagation. On the other side,
permanent effects are related to irreversible ground failure, induced
either by lateral spread/vertical settlements as a result of soil lique-
faction, or by lateral/oblique displacement as a result of slope in-
stability, or by horizontal/vertical offset due to fault rupture crossing
[2].
The experience from past earthquake has shown that, apart from
few cases (e.g. [1,2]) structural failures of pipelines are mainly asso-
ciated to permanent ground deformations (see, e.g., [3–5]). Among the
latter cases, pipelines are mostly vulnerable to active fault crossings
(e.g. [6]) since the strength of the surrounding soil may not reduce
significantly as in the liquefaction or landsliding problems, and loca-
lized large straining are likely to occur along the pipeline. Several well-
known examples of major pipeline failures induced by the interaction
with the seismic fault offset are from San Fernando (California, 1971)
[7], Manjil (Iran, 1990) [8], Kocaeli (Turkey, 1999) [9], and Chi-Chi
(Taiwan, 1999) [10].
Analytical modeling of the buried pipeline-fault interaction problem
started with the pioneering work of Newmark and Hall ([11,12]), who
modeled the pipeline as a cable connected to the soil with nonlinear
springs, subsequently refined by Kennedy et al. [13]. These works
provided the theoretical support to the 1984 ASCE Guidelines [14],
which have been for decades the basis for seismic analysis design of
pipelines crossing active faults. Subsequently, different researchers
provided improvements, in the framework of simplified analytical ap-
proaches, such as [15–20].
Only in the recent years, a substantial experimental activity pro-
vided support to the previous analytical approaches. Aside from the
shaking table tests performed by Sim et al. [21], who used a special
equipment, able to simultaneously simulate the transient vibration due
to seismic wave propagation together with the permanent ground dis-
placement caused by fault offset, research work was mostly based on
centrifuge tests, such as [3,22], and [23], who verified that the 1984
ASCE Guidelines [14] are in good agreement with the experimental
results for strike-slip faulting, but they tend to overestimate stiffness

⁎
Corresponding author at: Department of Civil and Environmental Engineering, Politecnico di Milano, Milano, Italy.
E-mail addresses: aliguney.ozcebe@polimi.it (A. Özcebe), roberto.paolucci@polimi.it (R. Paolucci), stefano.mariani@polimi.it (S. Mariani).

http://dx.doi.org/10.1016/j.soildyn.2017.07.017
Received 24 December 2016; Received in revised form 31 May 2017; Accepted 24 July 2017
0267-7261/ © 2017 Elsevier Ltd. All rights reserved.
A. Özcebe et al.
and strength of the p-y curves in normal faulting. Centrifuge tests on
reverse faulting were reported by Rohjani et al. [8], who noted the
dependence of the instability failure modes on the burial depth, moving
from general buckling, at shallow depths, to cross-sectional wrinkling,
at larger depths.
Both the analytical and experimental works have confirmed the
following factors that contribute to the safety of pipelines crossing fault
ruptures, (see e.g., Eurocode 8 Part 4 [24]): (i) the depth at which the
pipeline is buried should be minimized; (ii) the fault crossing angle of
the pipeline should be chosen to promote tensile elongation and reduce
compressive strains; (iii) the cross-sectional thickness to diameter ratio
should be increased; (iv) the angle of interface friction between the
pipeline and the soil should be reduced, by selecting a loose to medium
granular soil as backfill material.
Simplified analytical approaches proved to be a very effective tool
in understanding the key factors influencing the seismic response of
buried pipelines crossing fault ruptures, as well as in determining the
threshold values of fault offset to achieve initial pipe yielding, i.e., a
damage limit state. However, being typically based on modeling the
system as a linear beam on a bed of non-linear springs, they are not
suitable to follow extensively the response of the pipeline in the non-
linear range, within a performance-based design approach. As a matter
of fact, at the relatively large strain levels that are typically prescribed
to define the ultimate limit state for a buried pipeline subject to fault
rupture, i.e. 3% in tension and 1% in compression according to EC8
Part4, local buckling may occur, which can be reliably predicted only
by means of three-dimensional (3D) finite element (FE) numerical ap-
proaches.
However, the accurate numerical modeling of this complicated soil-
structure interaction problem poses several major numerical difficul-
ties, such as: i) 3D geometry; ii) large deformations; iii) local cross-
sectional buckling; iv) Eulerian buckling under compressive fault
movement; v) pipe sliding with respect to the surrounding soil; vi) non-
linear soil behavior. As a result, such comprehensive numerical ana-
lyses are still confined at the research stage, and relatively few ex-
amples exist of 3D simulations involving the large-strain non-linear
response of all the elements of the problem, i.e., the structure, the soil,
and the soil-structure interface (e.g. [4,18,25–29]). In some research
works (e.g. [23] and [30]), the complexity of the problem is reduced by
considering a 3D structural model on a bed of non-linear springs, but, in
the latter case, the contribution of interaction among the different
components of loading at the soil-foundation interface is lost.
In this paper, we will address, through a 3D numerical finite ele-
ment approach involving discretization of both soil and pipeline, the
response of pressurized buried pipelines subjected to normal fault
rupture, which is the common case in Central/Southern Italy, with
expected maximum magnitude around 7, corresponding to a maximum
expected fault-offset of 1–1.5 m [31]. Compared to previous studies on
the interaction of normal fault offsets with buried pipelines (e.g. [17]
and [23]), the contribution of this paper will mainly focus on the fol-
lowing objectives:
– to highlight the main computational problems to be tackled, espe-
cially as regards the longitudinal extent of the model to avoid
spurious effects from boundary conditions, that in such computa-
tionally demanding 3D numerical simulations may be a critical
issue;
– to establish relationships between the pipe performance, measured
through standard definitions of serviceability and ultimate limit
states, with increasing values of the fault offset;
– to highlight the role, within such relationships, of several para-
meters, related to geometry (dip angle, fault-pipe intersection
angle), to mechanical properties (soil elastic modulus, interface
friction angle), and to operating conditions (internal pressure and
temperature of the pipe);
– to analyze the changing shape and extension of the plastic zones
106
Soil Dynamics and Earthquake Engineering 101 (2017) 105–115
within the pipe as a function of load conditions;
– to build a set of reliable numerical results to be used as a benchmark
to develop and calibrate more simplified approaches usable in the
engineering practice.
2. Numerical modeling
To provide reliable results, the adopted FE approach should prop-
erly account on one side for the loading conditions, in terms of oper-
ating gas pressure and temperature, of the dead load due to the burying
soil layer, as well as of the fault geometry and kinematics, in terms of
intersection angle and direction and amount of the fault offset. On the
other side, it should properly model the large amount of nonlinear re-
sponse involved in the pipe and soil responses, including that at their
interface.
As a first step, we defined a region including the pipeline and the
surrounding soil, the dimensions of which were set in order to avoid
disturbances of the boundary conditions on the structural response.
This is of concern especially in the longitudinal direction of the pipe,
because it should be checked that, in the most heavily strained region
close to the fault crossing, the solution is not affected by the numerical
boundary conditions which limit the longitudinal extension of the
model. As illustrated later, such extension depends on the combination
of pipe geometry, non-linear soil response, soil-pipe interaction, fault
dip angle and maximum offset, so that it cannot be set a-priori in the
analysis, even if its actual impact should be verified a-posteriori.
When symmetry of loading conditions applied along the vertical
longitudinal plane holds, only half of the full model was considered in
the analysis and discretized.
The loading conditions are applied in two steps. First, the operating
conditions in terms of gas pressure and temperature are provided at the
inner surface of the pipe, together with the gravity load caused by the
soil burying the pipe, leading to a small ovalization of the cross-section.
These loads were applied quasi-statically, and provide the initial con-
ditions from which the fault offset is plugged in the model through
progressively increasing values. Due to the prevailing effect of large
permanent values of fault offset on the pipeline response, the con-
tribution of transient loading during the seismic excitation, including
inertial effects, was disregarded.
The mesh size was set to cope with the two competing requirements
of accuracy and minimization of computational costs. Preliminary
analyses were run to design an optimum space discretization, to ensure
the structural response within the region of interest to be mesh-in-
dependent. A finer mesh discretization was selected close to the fault
crossing. To avoid numerical locking when large values of the fault
offset are applied and large plastic deformations occur, the pipe and the
soil were discretized through reduced-integration (with hourglass
control) 4-node shell elements and 8-node solid elements, respectively.
Numerical analyses were carried out with the commercial FE code
Abaqus [32].
3. Numerical results
3.1. Validation against centrifuge test results of normal fault rupture
propagation
Bransby et al. [33] conducted an experimental campaign, based on a
centrifuge test carried out to investigate the propagation of a normal
fault inside dry, medium dense (Dr = 60%), uniform silica sand with
the following properties: 50%-passing diameter (d50) = 0.3 mm; max-
imum dry density (ρmax) = 1.7 t/m3; minimum dry density (ρmin) =
1.44 t/m3; internal effective friction angle (ϕ’) = 35° and dilation angle
(ψ) = 6°. The physical model set-up is shown in Fig. 1, with a geometry
scaling factor of 115. The main output of the test was the measurement
of the displacement field, from which the shear strain field was then
computed.
A. Özcebe et al. Soil Dynamics and Earthquake Engineering 101 (2017) 105–115

Fig. 1. Drawing of the physical model at the prototype scale (after [33]).

In the numerical counterpart considered here, still at the prototype

scale, plane strain conditions were assumed. The sand response was
modeled by an isotropic, elastic perfectly-plastic constitutive law, with
Fig. 3. Comparison between experimental (after [33]) and numerical vertical displace-
equivalent shear modulus (Geq) varying between 20–30 MPa; Poisson's
ment profiles at a depth from surface z = − 0.9 m. Results obtained using different
ratio νs = 0.25–0.30, and yielding governed by a Mohr-Coulomb failure amplitudes Δx (= 1 m, 0.5 m, or 0.25 m) of the region across which the fault offset is
condition with non-associated flow rule. The non-uniformity of the imposed along the bottom surface of the model are compared.
gravitational field resulting from the centrifugal acceleration and
Earth's gravity acceleration was not taken into consideration. The fault
bottom surface of the model, showing a limited dependency of the so-
offset was applied as a rigid body motion on the boundaries of the
lution using widths smaller than 1 m.
hanging wall (HW) side, as shown in Fig. 1. To avoid unrealistically
An acceptable agreement between the experimental and numerical
large strains at the interface with the footwall (FW) boundary, a narrow
displacement profiles was obtained. The formation and inclination of
1 m wide region, in between the HW and the FW, was introduced for
the main shear band in the sand was also satisfactorily predicted, as
the fault offset to gradually increase from zero to its maximum value.
well as the width of the main shear plane close to fault outcropping, as
A mesh sensitivity analysis was preliminary carried out, adopting a
shown in Fig. 3 by the spreading of the region wherein the vertical
gradual refinement from a coarser mesh away from the fault crossing
displacement jumps. The mismatch between the numerical and ex-
(with a characteristic element size of 1.0 m) to a finer one close to the
perimental datasets may be related to having neglected some effects
fault (with a characteristic size of 0.25 m). By comparing experimental
like: non-uniform gravitational field at centrifuge conditions, softening
data and numerical predictions, it was possible to point out that the
of sand, strain localization after yielding, inhomogeneous soil condi-
discretization along the horizontal direction plays the most important
tions.
role; accordingly, a structured mesh of quadrilateral elements featuring
It is finally noted that, to pursue the main objectives of this study,
node spacing of 0.25 m and 0.50 m in the horizontal and vertical di-
i.e., investigating the pipeline performance up to the serviceability and
rections, respectively, provided a reasonable trade-off between accu-
ultimate limit states under increasing values of fault offset, an extreme
racy and computational efficiency.
mesh refinement to capture very localized solutions of soil failure was
In Fig. 2, displacement and strain fields obtained with such space
not considered necessary. A similar compromise between accuracy of
discretization are compared with the available experimental data for a
the solution and objectives of the analysis is also at the base of the
fault offset of around 1.2 m, see [33]. The corresponding vertical dis-
selection of the structural grid size, which is addressed in the next
placement profiles at a depth of z = 0.9 m, are compared in Fig. 3. In
section.
this figure, numerical results are also illustrated at varying amplitudes
Δx of the region across which the fault offset is imposed along the

Fig. 2. Comparison between experimental (top row, after [33]) and numerical (bottom row) results for the centrifuge model at a fault offset of around 1.0 m. Left column: displacement
field; right column: strain field.

107
A. Özcebe et al.
3.2. Case studies of pipe-fault rupture interaction
We start the set of numerical simulations by a benchmark case,
where a steep dip angle was associated to the fault plane (δ = 80°), and
the fault offset (ΔF) was increased up to a maximum value of 2 m. The
study was carried out by considering first the effect of the fault
offset alone (Case 1), then the combined loading by the internal gas
pressure and fault offset (pi = 8 MPa, Case 2) was applied, and, finally,
the operating conditions were completed by considering the additional
contribution of the variation of temperature (ΔT = 50 °C, Case 3). For
all of those cases, the simulation was repeated for lower values of the
dip angle (δ = 40°–60°), in order to verify its effect not only on the
resulting stresses and plastic deformations of the pipe, but also on the
geometric modeling assumptions, especially to check whether the
length of the modeled portion of the pipeline was sufficiently extended.
In addition to these benchmark numerical simulations, we con-
sidered other situations, in order to study the impact of different as-
sumptions, such as the reduced value of the elastic soil modulus (Case
4), the reduced shear strength, both in terms of the friction (φ‫ )׳‬and
dilatancy (ψ) angles (Case 5), the increase of the soil-pipe interface
friction coefficient, to investigate the increased constrain provided by
the soil to the pipe movement (Case 6). Finally, in Case 7 we re-con-
sidered the benchmark case with operating conditions (Case 3), and
checked the effect of a change of plane orientation, i.e., instead of
considering the unit vector normal to the fault plane directed as the
longitudinal axis of the pipe (αf = 0°), we selected αf = 45°. Therefore,
only in the latter case symmetry conditions could not be applied. The
list of case studies is reported in Table 1.
For all simulations, the geometric characteristics (cross-sectional
thickness and diameter) as well the pipeline burial depth were kept
constant. A shallowly embedded (H = 1.5 m), large diameter (D =
1.2 m) gas transmitting pipeline with wall thickness t = 0.02 m was
modeled. Trench depth (5 m) and width (5.2 m) were chosen con-
sistently with a real case study. Consistently with the former verifica-
tion stage, the mesh density for the soil is characterized by an average
element size in the radial direction of 50 cm. 32 shell elements were
used for the spatial discretization of the pipe along its perimeter. By
keeping an almost unitary shape ratio for the shell elements, their
characteristic size resulted to be around 0.16 m. Although such element
size along the longitudinal axis of the pipe is not sufficient to accurately
model the evolution of local buckling induced by compressive stress
states, it is refined enough to capture the triggering of this ultimate
limit state, beyond which our numerical simulations were terminated.
The ×60 graded pipe steel was modeled by an elasto-plastic con-
stitutive law, with yielding governed by von-Mises law (fy = 450 MPa),
Table 1
List of case studies with the parameters considered in the sensitivity analysis. Soil parameters: shear wave velocity (Vs), mass density (ρ), equivalent linear shear modulus (Geq), soil
Poisson's ratio (v), internal effective friction angle (φ’), dilation angle (ψ), effective cohesion (c’) Structure parameters: Modulus of elasticity for steel (Es), steel Poisson's ratio (νs), yield
strength of steel (fy), bilinear hardening ratio (r).
# Soil Structure
1 Vs=175 m/s, ρ=1.85 t/m3 Geq=0.33Gmax, Es = 210 GPa,
ν=0.25, ϕ’=36°, ψ’=6°, c’=0.1 kPa νs = 0.30
fy = 450 MPa,
r = 0.002
2 as case 1
3 as case 1
4 as case 1 with Geq = 0.15Gmax as case 1
5 Vs = 125 m/s, ρ = 1.60 t/m3 Geq = 0.33Gmax, as case 1
ν = 0.25, ϕ’ = 30°, ψ’ = 0°, c’=0.1 kPa
6 as case 1
7 as case 1
Soil Dynamics and Earthquake Engineering 101 (2017) 105–115
with associated flow rule, and featuring a slight isotropic hardening (r
= 0.002). The soil was instead assumed to display an elastic, perfectly-
plastic response governed by Mohr-Coulomb yield condition, with non-
associated flow rule. Further details of the numerical mesh along the
cross-sectional plane and of the prescribed material properties are
shown in Fig. 4.
A preliminary investigation was made to determine the optimum
longitudinal size of the numerical model, depending on the combined
effect of fault offset, dip angle and mechanical soil properties.
Eventually, for a single numerical model to be investigated, a con-
servative value L = 2000 m was chosen for all case studies. As shown in
Fig. 5, the density of the mesh was gradually decreased for increasing
distance from the fault rupture, after verification of the mesh in-
dependence of the numerical results.
Similar to the validation example of Section 3.1, boundary condi-
tions to model the fault slip were applied as follows. For the soil body:
(i) the whole outer surface (except for the top) of the FW part was fixed
in their outward normal directions and for bottom surface also hor-
izontal fixity is provided, (ii) bottom and vertical planes of the HW part
were moved with an increased value of displacement, and (iii) long-
itudinal sides of HW were fixed along their outward normal directions
throughout the longitudinal length of the model L. For the pipe ele-
ments: (iv) pipe cross-section on the extreme FW side was restrained
along the pipe longitudinal direction, (v) pipe cross-section on the ex-
treme HW side was moved with an increased value of displacement (as
in (ii)), and (vi) restraints according to symmetry conditions were
provided at the symmetry plane, all along L. In Fig. 6, in-plane
boundary conditions (i.e. (i), (ii), (iv), (v)) are shown.
The soil and pipe elements are connected with zero length interface
(contact) elements, for which no normal deformation is permitted,
while shear deformations are controlled by the shear strength of the
interface. The latter is taken as 2/3 of the soil effective friction angle, in
agreement with ALA Guidelines [34].
3.3. The benchmark case
To display the evolution of stress and strain throughout the pipeline
as a function of the amplitude of fault offset, in Fig. 7 the spatial var-
iation of the Mises stress, normalized by the yield stress, is shown, to-
gether with the longitudinal strain computed along the upper and lower
lines of the pipe for ΔF = 0.4–0.8 m–1.5 m.
Yielding is first attained in the FW side of the pipe along the upper
line, for a fault offset slightly larger than 0.4 m, and, as ΔF increases,
yielding appears also on the HW side, along the lower line. In both
cases, yielding occurs due to tensile stresses. At ΔF = 1.5 m, yielding is
Interface Fault angle Operating Aim
conditions
kϕ = 0.67 αf = 0 ° pi = 0 MPa – Benchmark case
– Effect of dip angle
ΔT = 0 °C
pi = 8 MPa Effect of internal gas pressure alone
ΔT = 0 °C
pi = 8 MPa Combined effect of internal gas
ΔT = 50 °C pressure and temperature difference
Effect of soil elastic moduli
kϕ = 0.90 as case 1 Effect of soil bearing pressure
(with equal interface friction)
kϕ = 1.00 as case 1 Effect of interface friction
(with equal soil bearing pressure)
αf=45° pi = 8 MPa Effect of fault crossing angle
ΔT = 50 °C
108
A. Özcebe et al.
largely diffused throughout a zone of about 50 m around the fault
crossing, dominated by bending, which is represented in more detail by
Fig. 8. Note that, while on the FW side the plastic deformation is con-
fined within a relatively narrow zone, roughly associated to the for-
mation of a plastic hinge, yielding on the HW side is distributed along a
much wider zone with Mises stress along the inner side of the pipe
which hardly reach yielding also for ΔF = 1.5 m.
Another important result to point out for this initial case study,
refers to the extension of the portion of the model affected by different
deformation modes. As previously noted, a zone with predominant
bending deformation appears across the fault intersection, with spatial
extension L2 ≈ 50 m. Externally, both on the FW and HW sides, a re-
gion of prevailing longitudinal elongation appears, the extension of
which, denoted by L1FW and L1HW (see sketch in Fig. 9), is related to the
amplitude of the fault offset. Although straining in this region provides
a limited contribution to the total deformation, the minimum long-
itudinal extension of the numerical model should include at least such
elongation zones, in order to avoid any spurious effect on the results
owing to numerical boundary conditions too close to the fault crossing.
For this reason, we provide in Fig. 10 quantitative evaluations of L2,
L1HW and L1FW. Results are also shown for the operating conditions,
with internal gas pressure pi = 8 MPa and temperature variation ΔT =
50 °C, which will be commented in more detail in the following section.
From the left and mid panels of Fig. 10, it can be deduced that the
spatial extension of L1FW and L1HW is about the same, irrespective on
the operating conditions, and it only depends on ΔF,x = ΔFcosδ, which
is the projection of the fault offset along the pipe axis. As regards L2, it
is nearly independent on the magnitude of ΔF, and roughly equals 50 m,
where about 20% of such extension is on the FW side, where the largest
plastic strains occur, while the remaining is on the HW side. Besides the
benchmark value δ = 80°, additional dip angles were considered (δ =
90°, δ = 60°, δ = 40°). Relevant results are presented in the perfor-
mance based framework in Table 2.
109
Soil Dynamics and Earthquake Engineering 101 (2017) 105–115
Fig. 4. Finite element mesh in the cross-sectional plane and ma-
terial properties for the benchmark model.
Fig. 5. Finite element mesh along the longitudinal direction for
the benchmark case.
3.4. Impact of operating conditions
We now focus on the variation of the solution when operating
conditions are also considered in addition to the fault offset loading. A
similar investigation was also reported in [20]. We started from con-
sidering the internal gas pressure alone (Case 2, pi = 8 MPa) and
subsequently a temperature variation ΔT = 50 °C was also introduced
(Case 3). In Fig. 11, the stress ratio (σM/σY) and the corresponding
longitudinal strain (εl) observed in the extreme top and extreme bottom
fibers are shown, for the single case of fault offset ΔF = 1.5 m. In ad-
dition, the cross-section ovalization is also shown.
If we compare Fig. 11 with the corresponding profiles for Case 1,
illustrated at the bottom of Fig. 7, some important remarks can be
made:
– Owing to the effect of pi, cross-sectional ovalization is significantly
reduced from about 3% in Case 1, to about 0.7% in both Cases 2 and
3;
– Due to the initial operating conditions, straining due to the fault
offset is larger, both in tension and compression, with significant,
albeit localized, plastic hinging on the FW side, but also with a more
widespread plastic zone along the HW side. The Mises stress pattern
for Case 3 is illustrated in Fig. 12.
– The length of the plastic hinge along the FW side extends sig-
nificantly, up to about 7 times the pipe diameter. To clarify this
result, we illustrate in Fig. 13 the evolution of the extent of the
plastic zone in tension and compression, as a function of the fault
offset and of the operating conditions. It appears that consideration
of ΔT has a major impact on the length of plastic zone in com-
pression, which becomes almost as wide as the plastic zone in ten-
sion, while in the other cases plastic hinge length is dominated by
tension.
Fig. 6. Boundary conditions at the second stage of the analysis
(after application of operating loading conditions), for the quasi-
static application of fault offset. Out-of-plane boundary conditions
are not shown.
A. Özcebe et al. Soil Dynamics and Earthquake Engineering 101 (2017) 105–115

Fig. 7. Case 1: Pipe response in terms of normalized Mises stress,

longitudinal strain and cross-section ovalization (for the most-
deformed cross-sections on footwall region). Fault offsets 0.4
(top), 0.8 (middle) and 1.5 m (bottom). The values for the upper
line of the pipe (thick dashed), and for the lower line (thin gray)
are shown. The undeformed (dashed line) and deformed (con-
tinuous line) cross-sections are also illustrated.

Fig. 8. Case 1: deformed configuration of the pipe for fault offset

1.5 m in the vicinity of the fault crossing, with identification of
the Mises stress distribution. In zoom-up windows, the plastic
equivalent strains are indicated (black color: PEEQ > 10-5).

Fig. 9. Case 1: deformed configuration of the pipe for fault offset

1.5 m, and identification of the main deformation zones. L1FW and
L1HW: length of the pipe trunk subject to tensional, along the foot
(FW) and hanging wall (HW), respectively. L2: length of the pipe
trunk subject to combined tension and bending. Panel denotes the
zoomed area illustrated in Fig. 8. MCPs: Maximum curvature
points.

110
A. Özcebe et al. Soil Dynamics and Earthquake Engineering 101 (2017) 105–115

Fig. 10. Effect of dip angle on the deformation features (see Fig. 9) of the pipeline, for Cases 1 (pi = 0, ΔT = 0), 2 (pi = 8 MPa, ΔT = 0) and 3: pi = 8 MPa, ΔT = 50 °C. Note, from the
left and middle panels, that the length of the pipeline trunk in tension is about the same for the FW and HW sides and depends on ΔF,x = ΔFcosδ, which is the projection of the fault
offset along the pipe axis. On the right side, it is shown that the length of the pipe trunk mainly subject to combined tension and bending (L2) has only a moderate dependence on ΔF, and
its value is around 50 m.

3.5. Pipeline performance
The pipeline performance for increasing fault offsets was verified
based on the following criteria, by making reference to Part 4 of
Eurocode 8 [5]:
Damage limit state (DLS): εt , y = εc, y = 0.002
bottom panel in Fig. 14). This is a somehow surprising result, because a
pipeline under normal faulting is expected to be mostly subjected to
tensile straining. However, the presence of a high variation of tem-
perature (ΔT = 50 °C) induces a significant compressive pre-stress
which adds to the compressive bending stresses along the FW side of the
pipe when fault offset occurs.
(1a)

Ultimate limit state (ULS): εt , u = 0.03
Ultimate limit state (ULS): εc, u = min( −0.01, −0.4t / D)
where, in Eq. (1a), εt,y and εc,y are the yield longitudinal strains, while
in Eqs. (1b) and (1c) εt,u and εc,u are the ULS strains in tension and
compression, respectively, t represents the cross-sectional thickness and
D is the pipe diameter. Owing to the distance between DLS and ULS
performance levels, an intermediate damage threshold limit state (ILS)
was also considered, represented by a strain value 0.5%.
Following ALA Guidelines [34], we also addressed the limit for
ovalization deformation (ranging from 2% to 5%), but, according to
results shown previously, this is not a matter for operating conditions
because of the positive effect of internal gas pressure.
Performance checks are shown in Fig. 14 for Cases 1, 2 and 3, re-
ferring to the fault dip angle δ = 80°. It is found that yielding first
occurs in compression in Case 3, for a fault offset of about 40 cm (see
(1b) 3.6. Effect of non-perpendicular fault crossing
(1c)
Since the orientation of the fault rupture plane may be subjected to
large uncertainties, also because of the complicated patterns based on
which the fault rupture may propagate through the shallow ground
surface layers, it is interesting to determine the sensitivity of the solu-
tion to the fault plane orientation, which, up to now, was considered to
have a unit normal vector directed as the longitudinal axis of the pipe
(αf = 0°, as in Fig. 15). A thorough sensitivity study implies a large
computational effort and it is beyond the scope of this paper. However,
to start shading light on this subject and to point out its practical re-
levance, we have considered a single case study, obtained based on
Case 3 (δ = 80°, pi = 8 MPa, ΔT = 50 °C), but with a normal vector
forming an angle αf = 45° with the longitudinal axis of the pipe, as
sketched representatively in Fig. 15.
The non-perpendicular fault-crossing condition turns out to be more

Table 2
Values of fault offset (in m) associated to the exceedance of limit states defined for the case studied. Damage limit state (DLS, ε = 0.2%); intermediate damage limit state (ILS, ε = 0.5%);
ultimate limit state (ULS, Eqs. 1b and 1c)); ovalization (OLS). Attainment of the limit state in compression or in tension is denoted by C or T, respectively. The values in parantheses are
the fault offset values at first yield obtained by using the simplified analytical procedure of Karamitros et al. [17], which are noted as their first exceedances for each 10 cm increase in the
fault offset.

DLS ILS ULS OLS

ΔF (m) T/C ΔF (m) T/C ΔF (m) T/C 2% 5%

Case 1-δ = 80° 0.4–0.5 T 0.9–1.0 T >2 – 0.8–0.9 >2

(0.6)
Case 1-δ = 90° 0.5–0.6 C 0.9–1.0 C 1.0–1.1 C 0.8–0.9 >2
Case 2-δ = 80° 0.4–0.5 T 0.8–0.9 T >2 – >2 >2
Case 3-δ = 80° 0.3–0.4 C 1.0–1.25 T 1.5–1.75 C > 1.75 > 1.75
Case 4-δ = 80° 0.5–0.6 T 1.0–1.25 T >2 – 1.0–1.25 >2
Case 5-δ = 80° 0.6–0.7 T 1.0–1.25 T >2 – 1.5–1.75 >2
Case 6-δ = 80° 0.3–0.4 T 0.7–0.8 T >2 – 0.8–0.9 >2
Case 1-δ = 60° 0.4–0.5 T 0.9–1.0 T >2 – 1.25–1.5 >2
(0.5)
Case 2-δ = 60° 0.4–0.5 T 0.8–0.9 T >2 – >2 >2
Case 3-δ = 60° 0.6–0.7 T 1.0–1.25 T >2 – >2 >2
Case 4-δ = 60° 0.5–0.6 T 1.0–1.25 T >2 – 1.5–1.75 >2
Case 5-δ = 60° 0.6–0.7 T 1.25–1.5 T >2 – >2 >2
Case 6-δ = 60° 0.3–0.4 T 0.5–0.6 T 1.25–1.50 T 1.0–1.25 1.25–1.5
Case 1-δ = 40° 0.5–0.6 T 1.0–1.25 T >2 – 1.75–2.0 >2
(0.5)
Case 2-δ = 40° 0.4–0.5 T 0.8–0.9 T >2 – >2 >2
Case 3-δ = 40° 0.7–0.8 T 1.0–1.25 T >2 – >2 >2
Case 4-δ = 40° 0.5–0.6 T 1.0–1.25 T >2 – 1.75–2.0 >2
Case 5-δ = 40° 0.6–0.7 T 1.25–1.5 T >2 – >2 >2
Case 6-δ = 40° 0.3–0.4 T 0.5–0.6 T 0.8–0.9 T 0.7–0.8 > 0.9
Case 7-δ = 80° 0.2–0.3 C 0.6–0.7 C 0.8–0.9 C > 0.9 > 0.9

111
A. Özcebe et al.
severe, with all limit states reached in compression, and limit fault
offsets of about 0.2–0.3 m for DLS (against 0.3–0.4 m for the perpen-
dicular case), 0.6–0.7 m for ILS (against 1.0–1.25 m, reached in tension
for the perpendicular case), and 0.8 m for ULS (against 1.5–1.75 m) (see
Table 2, Case III vs Case IX). To clarify reasons of such result, in Fig. 16
the longitudinal strain distribution is plotted along the upper and lower
generatrixes of the pipe, for the case of a fault offset ΔF = 1.0 m and αf
= 45°. Furthermore, the polar distribution of the longitudinal strain in
the most heavily loaded cross-sections on the FW and HW sides of the
pipe are shown.
It turns out that the pipe in the non-perpendicular case is negatively
affected by the following main factors: (a) given the same normal fault
offset, the elongation component is reduced and the bending compo-
nent increased, with a consequent increase of the compressive strain in
the most critical cross-section of the FW trunk; (b) mobilization of
passive soil pressure at the lateral side induces a further constrain
which tends to increase bending of the pipe, which reaches yielding
also within the HW trunk.
Although this is a preliminary result, which deserves further in-
vestigations and numerical simulations, it should be pointed out that
the common assumption of disregarding the non-perpendicular com-
ponents of normal faulting may lead to unconservative results. This is
especially true for those cases, like the pressurized pipe subject to
temperature variation considered in this study, where the limit state of
the pipe in the perpendicular fault-crossing is reached in compression.
When the limit state is reached in tension, then the non-perpendicular
fault crossing may improve performance, but only until compressive
strains become predominant.
112
Soil Dynamics and Earthquake Engineering 101 (2017) 105–115
Fig. 11. Stress ratio (σM/σY), longitudinal strain (εl) and cross-
section ovalization (for the most-deformed cross-section on foot-
wall plastic hinge region). for Case 2 (top, pi = 8 MPa, ΔT = 0)
and Case 3 (bottom, pi = 8 MPa, ΔT=50 °C) for the fault offset ΔF
= 1.5 m (bottom). Dashed lines indicate the values at yielding.
3.7. Parametric study
A parametric summary of results for the different cases considered
in this study is reported in Table 2, in terms of fault offset value ΔF
required to reach a given performance level. The following indications
may be drawn.
– In general, normal fault crossings create tensile plastic deforma-
tions. However, in the case of operating temperature variation,
compressive strains may be significant, especially for high dip an-
gles, and govern the attainment of performance limit state.
– The internal pressure does not significantly affect performance in
terms of strain limits (see case 2 vs case 1). However, the ovalization
performance is significantly enhanced.
– The increase of operating temperature variation makes the system
slightly more resilient in terms of tensile deformation and sig-
nificantly more vulnerable in terms of compressive deformation (see
case 3 vs case 1). Since it is usually coupled with internal pressure,
the ovalization performance is enhanced.
– Decrease in dip angle (Cases 1–6) does not significantly change the
performance in tension. It significantly improves the performance of
the pipe against compressive failure. Ovalization performance is
inversely influenced.
– Increase of bearing pressure and elastic stiffness (Cases 4 and 5 vs
case 1) slightly decreases the performance of the pipe in terms of
tensile, compressive, and ovalization deformations.
– Increase of interface strength (Case 6 vs Case 1) significantly reduces
the performance in terms of tension. The reverse happens in terms of
Fig. 12. Case 3 (combined effect of p and ΔT): deformed config-
uration of the pipe for fault offset 1.5 m, with identification of
Mises stress (black color indicates yielding). In zoom-up windows,
PEEQ distributions are shown (black color: PEEQ > 10−5).
A. Özcebe et al. Soil Dynamics and Earthquake Engineering 101 (2017) 105–115

Fig. 13. Length of plastic zone in tension (LpFW,t) and compres-

sion (LpFW,c) of the most deformed portion of the pipeline along
the FW. Values are normalized by the pipe diameter D. Different
operating conditions (Cases 1, 2, 3) are considered.

compressive strains. Since tearing off is the likely failure me-

chanism, cross sectional deformations are also increased.
– Fault offsets to reach the DLS are compared (see values in par-
entheses for case 1 in Table 2) with those computed by the simpli-
fied analytical method proposed by Karamitros et al. [17], showing
a reasonably good agreement.
– For a dip angle δ = 90°, which was considered only for Case 1, an
abrupt reduction of the fault offset to reach the ULS is found, con-
sistent with the vanishing contribution of the fault offset elongation
component along the pipe axis. Moreover, governing mechanism of
the pipeline is shifted to compressional type in all performance le-
vels. Although this case is of limited practical relevance, since no
normal faults is expected along a vertical plane, it points out the
relevance of the elongation component of the fauls offset in the
overall good performance of the pipeline subjected to normal
faulting.
– Finally, it is important to note that non-perpendicular normal fault Fig. 15. Sketch of the case of non-perpendicular normal fault-crossing.
crossing (Case 7 vs Case 1) decreases performance in terms of
compressive straining and improves in terms of tensile straining.
Ovalization deformations were not investigated for non-perpendi- conditions, with fault offset up to 2 m, which is usually considered as an
cular normal fault crossing case. upper threshold in the case of normal faulting.
It was found that pipeline initially yields (DLS) for values of fault
4. Conclusions offset of about 0.3–0.5 m, with worst conditions occurring in the case of
high values of the friction angle at the pipeline-soil interface, or for
In this paper we investigated from a numerical point of view the large temperature gradients, typically increasing compression stress
performance levels of a gas pipeline subjected to normal fault rupture, throughout the pipeline. For increasing values of the fault offset, the
for increasing values of the fault offset and considering also the influ- ultimate limit state (ULS), defined as in the Eurocode 8 Part 4 in terms
ence of the operating conditions, including the effect of internal gas of a 3% longitudinal strain in tension and ultimate limit in compression,
pressure and variation of temperature. is reached only in few cases, typically related to high interface friction
After providing some practical guidelines to properly select the angles and low values of the fault dip angle.
spatial extension of the numerical model in the longitudinal direction, While this in general confirms the good performance of pipelines
in order to reduce the effect of boundary conditions on the numerical under normal faulting, mainly because of prevailing tensional stresses,
results, we carried out a set of 3D nonlinear numerical simulations, so that, in presence of sufficiently loose soils in the pipe trench, the fault
useful to clarify the response of the pipeline up to extreme loading offset may exceed in most cases 2 m without reaching the ULS, an

Fig. 14. Pipeline performance measured, as a func-

tion of fault offset, in terms of longitudinal strain in
tension (thick continuous line), longitudinal strain in
compression (thin continuous line), cross-sectional
ovalization. From top to bottom: Case 1, Case 2, Case
3. Attainment of a limit state is denoted by an arrow.

113
A. Özcebe et al.
important outcome of this research work is the sensitivity of the solu-
tion to the case of non-perpendicular fault-crossing, i.e., when the
normal to the fault plane is not contained in the longitudinal plane of
the pipe. Contrary to the expectations, the non-perpendicular crossing
case turned out to be more severe, with all limit states reached in
compression, and limit fault offsets of about 0.2–0.3 m for DLS and
0.8 m for ULS, in the case αF = 45°. This was argued to be related to the
mobilization of passive soil pressure at the lateral side of the pipe effect,
with an increase of the bending component of deformation, and a
corresponding increase of the compressive strain in the most critical
cross-sections of the pipeline.
Acknowledgements
Authors are grateful to two anonymous reviewers, the comments of
which contributed to improve the clarity of the paper.
References
[1] Datta TK. Seismic response of buried pipelines: a state-of-the-art review. Nucl Eng
Des 1999;192:271–84.
[2] O’Rourke MJ, Liu X. Response of buried pipelines subject to earthquake effects.
Monograph series. Buffalo, New York: Multidisciplinary Center for Earthquake
Engineering Research (MCEER); 1999. p. 249.
[3] Ha D, Abdoun TH, O’Rourke MJ, Symans MD, O’Rourke TD, Palmer MC, Stewart
HE. Buried high-density polyethylene pipelines subjected to normal and strike-slip
faulting – a centrifuge investigation. Can Geotech J 2008;45:1733–42.
[4] Vazouras P, Karamanos SA, Dakoulas P. Mechanical behavior of buried steel pipes
crossing active strike-slip faults. Soil Dyn Earthq Eng 2012;41:164–80.
[5] Xie X, Symans MD, O’Rourke MJ, Abdoun TH, O’Rourke TD, Palmer MC, Stewart
HE. Numerical modeling of buried HDPE pipelines subjected to strike-slip faulting. J
Earthq Eng 2011;15(8):1273–96. http://dx.doi.org/10.1080/13632469.2011.
569052.
[6] Moradi M, Rohjani M, Galandarzadeh A, Takada S. Centrifuge modeling of buried
continuous pipelines subjected to normal faulting. Earthq Eng Eng Vib
2013;12(1):155–64. http://dx.doi.org/10.1007/s11803-013-0159-z.
[7] O’Rourke TD, Palmer MC. Earthquake performance of gas transmission pipelines.
Fifth U.S.-Japan workshop on earthquake resistant design of lifeline facilities and
countermeasures against soil liquefaction, Snowbird, Utah. Technical report
NCEER-94-0026. Buffalo, New York: Multidisciplinary Center for Earthquake
Engineering Research (MCEER), A National Center of Excellence in Advanced
Technology Applications; 1994. 679–702.
[8] Rohjani M, Moradi M, Galandarzadeh A, Takada S. Centrifuge modeling of buried
continuous pipelines subjected to reverse faulting. Can Geotech J 2012;49:659–70.
http://dx.doi.org/10.1139/T2012-022.
[9] Eidinger JM, O’Rourke MJ, Bachhuber J. Performance of a pipeline under large fault
displacement. In: Proceedings of the 7th U.S. National conference on earthquake
engineering. Oakland, California; 2002.
[10] Miyajima M, Hashimoto T . Damage to water supply system and surface rupture due
to fault movement during the 1999 Ji-Ji earthquake in Taiwan. In: Proceedings of
114
Soil Dynamics and Earthquake Engineering 101 (2017) 105–115
Fig. 16. Top: Envelope of the maximum longitudinal strain dis-
tribution along the upper and lower generatrixes of the pipe.
Bottom: Polar plots of the longitudinal strain across the most
heavily loaded cross-sections on the FW (left) and HW (right)
sides of the pipe.
the 4th International conference Rec. Adv. in Geotechnical Earthq. Eng. and Soil
Dyn. San Diego, CA, Paper 10.45; 2001.
[11] Newmark NM, Hall WJ. Pipeline design to resist large fault displacement. In:
Proceeings of the U.S. National conference on earthquake engineering. Ann Arbor,
Michigan; 1975. 416-425.
[12] Hall W, Newmark N . Seismic design criteria for pipelines and facilities. Technical
Council of Lifeline earthquake engineering specialty conference. Los Angeles, NY:
ASCE; 31 August 1977. 18–34.
[13] Kennedy RP, Williamson RA, Chow AW. Fault movement effects on buried oil pi-
peline. J Transpor Eng Div-ASCE 1977;103(TE5):617–33.
[14] American Society of Civil Engineers (ASCE). Guidelines for the seismic design of oil
and gas pipeline systems. Prepared by the ASCE technical council on lifeline
earthquake engineering, D. Nyman Principal Investigator; 1984.
[15] Wang LRL, Yeh Y. A refined seismic analysis and design of buried pipeline for fault
movement. Earthq Eng Struct Dyn 1985;13:75–96.
[16] Karamitros DK, Bouckovalas GD, Kouretzis GP. Stress analysis of buried pipelines at
strike-slip fault crossings. Soil Dyn Earthq Eng 2007;27:200–11.
[17] Karamitros DK, Bouckovalas GD, Kouretzis GP, Gkesouli V. An analytical method
for strength verification of buried pipelines at normal fault crossings. Soil Dyn
Earthq Eng 2011;31:1452–64.
[18] Paolucci R, Griffini S, Mariani S. Simplified modeling of continuous buried pipelines
subject to earthquake fault rupture. Earthq Struct 2010;1(3):253–67. http://dx.doi.
org/10.12989/eas2010.1.3.253.
[19] Trifonov OV, Cherniy VP. A semi-analytical approach to a nonlinear stress-strain
analysis of buried steel pipelines crossing active faults. Soil Dyn Earthq Eng
2010;30:1298–308.
[20] Trifonov OV, Cherniy VP. Elastoplastic stress-strain analysis of buried steel pipe-
lines subjected to fault displacements with account for service loads. Soil Dyn
Earthq Eng 2012;33:54–62.
[21] Sim WW, Towhata I, Yamada S, Moinet GJM. Shaking table tests modeling small
diameter pipes crossing a vertical fault. Soil Dyn Earthq Eng 2012;35:59–71.
[22] Abdoun TH, Ha D, O’Rourke MJ, Symans MD, O’Rourke T, Palmer MC, Stewart HE.
Factors influencing the behavior of buried pipelines subjected to earthquake
faulting. Soil Dyn Earthq Eng 2009;29:415–27. http://dx.doi.org/10.1016/j.
soildyn.2008.04.006.
[23] Xie X, Symans MD, O’Rourke MJ, Abdoun TH, O’Rourke TD, Palmer MC, Stewart
HE. Numerical modeling of buried HDPE pipelines subjected to normal faulting: a
case study. Earthq Spectra 2013;29(2):609–32. http://dx.doi.org/10.1193/1.
4000137.
[24] CEN, European Committee for Standardization. Eurocode 8 – design of structures
for earthquake resistance, Part 4: silos, tanks and pipelines. Final draft; January
2006, prEN 1998-4.
[25] Joshi S, Prashant A, Arghya D, Jain SK. Analysis of buried pipelines subjected to
reverse fault motion. Soil Dyn Earthq Eng 2011;31:930–40.
[26] Vazouras P, Dakoulas P, Karamanos SA. Pipe-soil interaction and pipeline perfor-
mance under strike-slip fault movements. Soil Dyn Earthq Eng 2015;72:48–65.
[27] Tarinejad R, Mahdavi A, Jahangir N. Buried pipeline response analysis to reverse-
slip fault displacements. In: Proceedings of the 15th world conference on earth-
quake engineering. Lisbon, Portugal; 2012.
[28] Trifonov OV. Numerical stress-strain analysis of buried steel pipelines crossing ac-
tive strike-slip faults with an emphasis on fault modeling aspects. J Pipeline Syst
Eng Pract 2015;6:1.
[29] Özcebe AG, Paolucci R, Mariani S, Santoro D A numerical study of the pressurized
gas pipeline-normal fault interaction problem. In: Proceedings of the 6th interna-
tional conference on earthquake geotechnical engineering. Christchurch, New
Zealand; 1–4 November 2015.
A. Özcebe et al.
[30] Liu X, Zhang H, Li M, Xia M, Zheng W, Wu K, Han Y. Effects of steel properties on
the local buckling response of high strength pipelines subjected to reverse faulting.
J Nat Gas Sci Eng 2016;33:378–87.
[31] Wells DL, Coppersmith KJ. New empirical relationships among magnitude, rupture
length, rupture width, rupture area, and surface displacement. Bull Seismol Soc Am
1994;84(4):974–1002.
115
Soil Dynamics and Earthquake Engineering 101 (2017) 105–115
[32] Dassault Systèmes. Abaqus 6.13. Finite element code and the documentation; 2013.
[33] Bransby MF, Davies MCR, El Nahas A. Centrifuge modeling of normal fault-foun-
dation interaction. Bull Earthq Eng 2008;33(6):585–605.
[34] American Lifelines Alliance (ALA). Guidelines for the design of buried steel pipe.
American Lifelines Alliance; July 2001. Website: 〈www.americanlifelinealliance.
org〉.